Most social media is part of what is called the dark web.  This means that most is not indexed by traditional search engines. You will have to use specialized search engines.

I think you should not ignore google search or yahoo search as they could provide some social media data, but should rather focus on other search engines.

You could use a search engine called Twingly.

It is not free but they could offer you a six months free trial. They provide blog and forum data (forums are also considered social media).

Another tool is Webhose.io which offer forum conversations. They offer 2000 free searches per month. This is a very powerful tool from my experience.

You could also use the Twitter search API and the facebook graph API for traditional social media. The facebook API provides limited data, but you could get posts from fan pages and  groups.

https://www.google.fr/search?q=Twingly&ie=utf-8&oe=utf-8&gws_rd=cr&ei=SIZNVuqCOYPtaLeWvZgD

https://webhose.io/

https://dev.twitter.com/streaming/overview

https://developers.facebook.com/docs/graph-api

It is a good beginning. The leads it gives you can be further developed. 

I concur with responses to date. Moreover, Google Scholar appears to be increasingly recognized as a viable search engine. Overall, an eclectic approach to searching the literature so as to be interdisciplinary in scope as possible, is always recommended. A Masters and a PhD thesis often requires this. Always best to know what your supervisor and/or PhD committee thinks as well as they/she/he will be reviewing your work in the end.

Thanks everyone for sharing your inputs! 

I have utilized Factiva software for analyzing social media sentiments and planning to combine it with Google search as both sources appear to produce generally comparable (and adequate) measures of corruption trends.

Please do take a look at my new piece uploaded on applying a new attention cycle model to corruption (technical report) and comment!